We plan to study assaults committed by psychiatric inpatients using a closed-circuit television system now in operation on a special ward for the treatment of violent patients. The project has two main goals: 1) Characterize, in detail, assaults on the unit. Variables to be studied include assault frequency and seriousness, ward conditions (e.g. presence of structured activity), characteristics of the assailant and victim (e.g. diagnosis and psychopathology) and interpersonal cues. Cues are defined as warnings given by assailant and provocations emitted by the victim. These have never been accurately studied before. 2) Retrospectively calculate the likelihood of assault and of victimization given certain ward conditions, patient characteristics and interpersonal cues. Hypotheses to be tested include: 1) For the total group of patients certain cues will be sensitive and specific to assaults; 2) For some individual assaulters or victims certain cues will be sensitive and specific predictors of assault. Cues will be more specific as a function of characteristics of the other member of the assailant-victim dyad; 3) Assault is more likely during early morning hours, on a noisier ward, with fewer staff members present, and during unstructured activities. Methods. All episodes of assault will be saved on videotape for review. Control epochs Methods. All episodes of assault will be saved on videotape of cues. Ward conditions during assaults, including presence or absence of structured activities, number of staff members present and ward noise level will also be noted. For specificity determinations these variables will also be noted during review of videotape not recording assaults. Patient data to be collected includes BPRS scores every week, DSM-III-R diagnosis and neurologic impairments. This project may yield strategies for the prevention of assaults by psychiatric inpatients. Knowing retrospectively the likelihood of assault given certain environmental conditions, and certain interpersonal cues, environmental interventions can be planned. Staff can be trained to detect useful cues and patients can be trained to avoid provocative behavior.